The Growth Potential and Use Cases of both the Arweave and AO Ecosystems (NFA)

How does Arweave Work?

Arweave 1.0: Decentralised Permanent Data Storage

Arweave is a blockchain-based storage platform that provides permanent decentralised storage for users. Arweave is a protocol for creating the “Permaweb”, a web where information, data and applications are stored permanently and in a decentralised manner. Unlike traditional web storage services where content is subject to censorship, the Permaweb aims to create a permanent and censorship-proof repository for data. The original mission of Arweave was to develop a ‘Permaweb’ by permanently archiving all information published on the internet with a single upfront payment. However, the future use cases of Arweave will revolve around ensuring the permanent decentralisation and immutability of NFT’s, a market that is expected to triple in size over the next 5 years.

How it works:

Arweave utilises a unique data structure called blockweave, which is an iteration of the traditional blockchain model that can enable the permanent storage of large datasets. Blockchain’s traditional method of sequential confirmation connects every piece of data in a long singular verifiable chain. Each block in the blockweave is linked to the two previous blocks, the immediately preceding block and a randomly selected block from the network’s history (recall block). This dramatically reduces the amount of energy needed to run the blockweave as energy is not wasted verifying the same data repetitively.

Arweave’s mining system revolves around storing and replicating data on the network to maintain its permanence and accessibility. Miners earn rewards in AR tokens, Arweave’s native cryptocurrency for storing the data effectively. Arweave’s consensus mechanism, Succinct Proofs of Random Access (SPoRA), requires miners to provide cryptographic proof that they can access the recall block associated with the block they want to mine. This incentivises miners to store a larger amount of data, increasing the odds of having access to the random recall block and earning the rewards. Correspondingly, miners are also incentivised to keep rarer blocks instead of commonly replicated ones. While the previously linked block more likely includes a data block that is frequently accessed, the recall block incentivises miners to hold rarer blocks that have less miner competition in the equal likelihood they can still be chosen.

Arweave Tokenomics and Endowment:

Arweave utilises a native token AR, which it uses to reward miners and is paid for by users. The maximum supply of AR is 66 million with 55 million tokens minted at the network launch and the remaining 11 million to be introduced into circulation as block rewards.

The long-term sustainability of the Arweave network is dependent on the rewards earned for each block mined being greater than the combined operational costs for each block mined. The total mining rewards are the sum of all rewards miners receive for mining a block, which includes transaction fees and inflation rewards (which head towards zero as more supply is unlocked). If the sum of transaction fees and inflation rewards does not cover the operational/data storage costs then any additional remainder is taken from Arweave’s endowment fund. As of May 2024, this endowment fund has never been touched.

The network calculated the costs of maintaining the data stored in each block using a price per gigabyte per block metric (PGBB). This cost reflects the roughly 30.5% YOY decline in storage costs reflecting the Kyrder rate for Hard disk drives and Moore’s Law for solid state drives. People who store the permanent data on their computers are compensated in AR. The single payment that users put through are divided into two ways, as an upfront payment for the initial 200 years of storage (approx 20% of fees paid) and an endowment for future storage years (approx 80% of fees paid). Arweave uses an endowment model where the fees paid for storing data are invested in a manner that the interest generated can continue to fund the storage of that data forever. This model relies on the economic stability and utility of the AR token, creating a continual demand cycle as more investments into the token are required to sustain and grow the endowment.

Arweave’s working assumption is that the price of storage will continue to decline over time. Given technological improvements to be made and societies increasing need for data storage, this is a fair assumption. Over the past 50 years, data storage costs have gone down by an average of 30.5% per year. Arweave only assumes data storage costs will decline by 0.5% per year and any greater decrease than 0.5% adds to the number of years that data will be stored. It is important to note that Arweave cannot actually store data forever, however due to its untouched endowment it will likely store data for several life times. While Arweave cannot store data forever, no other blockchain or Web2 competitor can do the same.

Furthermore, the locking of AR tokens within the endowment fund is an effective burn of supply which provides strong deflationary pressure on the token. Roughly 80% of fees paid go into the endowment, meaning whenever someone uses Arweave for storage they are taking 80% of their spent tokens out of circulation for at least 200 years. This indicates that even though approximately 20% of the supply is still yet to be unlocked, this inflation is over compensated by tokens flowing into the endowment.

Ultimately, the largest risk this endowment model faces is that data storage costs decline by less than 0.5% per year or that the endowment is invested poorly. There is not substantial information on how the endowment is invested.

Utility and Market Growth:

Arweave distinguishes itself from its decentralised competitors by offering permanence, a critical feature that will drive its demand in the sector of NFT’s. Currently, the NFT market predominantly relies on centralised servers for storing actual NFT data due to the immense costs associated with on-chain storage. The most common scenario are blockchain platforms storing a URL pointing to the NFT’s location while the actual NFT is stored on a centralised server, which does not guarantee immutability. Centralised servers are subject to government oversight and potential government intervention, which can raise significant concerns for owners who value the long term security and accessibility of their NFT’s.

This vulnerability in the NFT market presents a substantial opportunity for Arweave to capitalise by providing immutable and permanent storage for a one-time only cost. This is a superior solution to centralised storage competitors as they are more secure, they have less risk of service disruption and are more cost efficient in the long term (see below).

As mentioned earlier, Arweave can be used for numerous use cases besides the storage of NFT’s such as archiving historical documents or data at risk of censorship. However, the overwhelming demand will likely come from NFT’s, a market expected to grow from its current valuation of $71.84 billion to an estimated $211.7 billion by 2030. The economic potential of NFT’s well into the future is significant, with use cases from gaming to virtual economics to the monetisation of digital culture. This projected growth reinforces the need for reliable decentralised storage solutions where Arweave is a market leader, providing them with an opportunity to act as a storage component for digital assets that cannot risk centralisation.

Costs:

The price of storage on Arweave has only recently become significantly more expensive than its decentralised competitors due to rapid appreciation of the AR token, however this is not the only influence. Arweave is designed to have a dynamic price for storage that is pegged to USD per gigabyte and the difficulty of the network. Arweave is maintained by a large decentralised network that must store the data and remain profitable. Data storage incurs additional expenses like cost of hardware and electricity which must be accounted for through the dynamic price to ensure miners remain profitable. Since the introduction of dynamic pricing in March 2023 the price of storing one gigabyte of data has risen from roughly 0.15 AR/GIB to 0.75 AR/GIB now, a 500% increase. This rate is expected to fall by ardrive.io but there is a strong possibility this does not occur, especially if AR continues to appreciate.

However, this still makes AR cheaper to store than the largest Web2 storage providers even though it is significantly more expensive than Filecoin. As mentioned earlier, the utility that Arweave offers is permanence and this is worth a premium but it is still possible that NFT owners prefer a substantially cheaper monthly cost to store their NFT’s rather than Arweave’s one time upfront fee.

Risks:

1. Competitors that can replicate Arweave’s model for a cheaper upfront price (such as NFT storage). However, none have been yet to do so with as conservative an endowment model or as long a guaranteed storage period.
2. NFT owners prefer a cheaper subscription based model for decentralised data storage rather than the permanent model offered by Arweave.
3. Miscalculation of future storage costs leading to a depletion of the endowment.
4. Unforeseen security threats that cripple the network.

While Arweave can be considered a pioneer as an individual storage layer, its utility extends far beyond providing permanent decentralised storage. As we will explore next, the introduction of a decentralised scalable computing layer built on top of Arweave will significantly enhance its potential in solving problems relating to AI and Automated Finance.

How does AO Work?

Arweave 2.0 — AO: Decentralised Scalable Compute

AO is a computing environment that operates on top of the Arweave network. Similar to other computing protocols like Ethereum or Solana, AO allows for the application of smart contracts, but differentiates itself by designing unlimited processes to run simultaneously in parallel instead of sequentially, allowing for unlimited scalability in theory. With AO, Arweave can be categorised alongside other L1’s, acting as a state layer for decentralised computing. AO also differentiates itself from other decentralised computing protocols like Akash by providing verifiability, achieved through perpetual storage on its base layer Arweave. By leveraging the verifiability and permanence of Arweave’s base layer, AO offers unique value by providing high-speed, low latency compute with unlimited scalability that is also verifiable. This is crucial for handling large datasets and complex computations, necessary for the development and application of artificial intelligence, an industry expected to grow to 2.575 trillion USD by 2032.

How does it work:

Firstly, let’s delve into computation protocols like Ethereum:

In Ethereum, every single user is sharing the same single thread of execution while globally locking the memory space when working. This means users have to work one at a time, in a massive global queue. Some architectures like rollups allow for more threads of execution to be executed at the same time, however this is only a partial improvement. This is because each thread is now hosting many contracts instead of individual ones but each community base is still competing to access a shared resource.

How does AO Work:

“AO follows the route internet services used to scale, not by improving the speed of individual computers but by harnessing many computers to work together in parallel to serve any number of users simultaneously” — Sam Williams.

In a queue of 1000 people, horizontal scaling can be thought of as spreading 1000 people into 1000 different queues while vertical scaling is improving the speed at which the single queue moves. AO utilises horizontal scaling by enabling any number of processes (individual computations) to occur at the same time and allowing each process to coordinate through messenger units. All of these processes are then permanently stored on AO’s base layer Arweave for verifiability.

Here’s the process:

1. When messages first enter AO, the user sends them to a messenger unit after paying in the currency they find most convenient, whether it be fiat or crypto. The messenger unit is responsible for finding the right scheduler unit in the distributed scheduler unit network.
2. The scheduler unit, which can take any form (decentralised or centralised) gives a sequence number to each message and ensures it is uploaded to Arweave directly.
3. The computing unit then computes the operation after the message is uploaded to Arweave and has reached consensus.
4. This result is sent back to the message unit, with the output of the message returned to the user. If the result is another computation that needs to be calculated then the process loops again.

Verifiability:

AO achieves its verifiability through its holographic state system, a fancy term for a permanent log of interactions on its base layer Arweave. This method creates a “hologram” of the state that can be independently verified by any third party.

Architecture of AO:

Theoretically, AO can offer unlimited computational power through the horizontal expansion of nodes (as long as there are sufficient network incentives). The three subnets refer to the three different network units responsible for message sorting, scheduling and computing. These can briefly be explained below:

• Processes: Individual units of computation that are executed within AO. Each process is stored permanently on Arweave’s message log and can thus be viewed in a “holographic state”. Each process can select its computing environment and thus processes can work independently from each other.
• Messages: Each interaction between processes are represented by messages and are sent between subnets.
• Scheduler Unit: A scheduler can be thought of as a more complex version of a sequencer, which orders transactions in a decentralised environment. In contrast, the scheduler unit in AO can take various forms (centralised or decentralised) and can even be hosted on different blockchains such as Bitcoin, allowing for increased resilience, customisation and interoperability. The scheduler unit gives a sequence number to each message and ensures it is uploaded to Arweave directly.
• Computing Unit: Compete in a peer-peer marketplace where the demand for the service pushes up the incentive to compute while competition pushes the price down for users.

What is fundamentally special about AO’s architecture is the separation of the consensus mechanism (on Arweave) from the computation on AO. Computation involves executing the code associated with smart contracts. In most traditional blockchains this computation is often directly tied to the consensus mechanism which means each node in the network must perform identical computations to ensure they get the same result.

In AO, the computation is decoupled from the consensus mechanism which means that the network can reach consensus on the state of the ledger without requiring each node to perform the same computations. This setup allows AO to significantly enhance their scalability unlike Ethereum, Solana and other L1’s, where network congestion leads to increased transaction fees. In contrast, AO’s scalable designs allow additional computation units to be added, maintaining high throughput while keeping transaction fees low. The absence of a traditional fee market means transaction costs on AO are not volatile, an attractive feature for small to large scale applications where predictable costs are needed for budgetary and operational planning.

Furthermore, AO is designed as a Single System Image (SSI) environment, where processes (units of computation), run independently and interact seamlessly through message passing. This eliminates bottlenecks that protocols like Solana face due to the use of shared memory. Solana and most other computing protocols use a shared memory model where all contracts can directly access and modify the memory of others synchronously. This design eventually encounters scalability barriers due to lock contention, which is where access to the same data by multiple users creates a bottleneck. AO’s message passing not only increases throughput but prevents the need for users to compete for system resources which provides an attractive stable cost for developers.

EVM and Scheduler Compatibility:

As mentioned earlier, AO allows developers to separate the requirements of computation which is handled by VM’s and transaction ordering which is handled by schedulers. Previously, developers would have to inherit the weakness of the VM they would be forced to use from the network they operate on. However, AO ensures that developers can improve computational efficiency and heighten security by incorporating custom operations suited to developer needs. This flexibility supports innovation and is another clear advantage compared to other computation platforms, attracting a diverse range of developers.

Here are two example use cases that would benefit from AO’s flexibility of VM’s:

1. If a developer wants to build a high-performance DEX that can handle a large volume of transactions and complex algorithms with specialised computational logic without sacrificing security or decentralisation:

Developers can choose a virtual machine that is optimised for high throughput and low latency financial transactions. This would be able to efficiently process and validate complex trading algorithms and financial models.

2. If a developer wants to develop a DApp designed for processing and developing LLM’s, they can select a VM optimised for machine learning tasks that supports GPU based computations which other blockchain VM’s likely lack.

Furthermore, a custom scheduler can prioritise tasks based on urgency and computational intensity. This means that a chosen scheduler can manage resource allocation more effectively by assigning computational resources that may be underutilised.

This is a clear differentiator that most other decentralised computation platforms do not provide. However, it is important to note that while AO is superior in customisability and verifiability to most computation platforms, due to the cost of storing each computation permanently the primary use cases for AO will be DApps that require both heavy computation and verifiability together.

Arweave Team

Arweave was founded by Sam Williams and William Jones and a DAO guides Arweave for community-led decisions and voting. The Arweave team is led by Sam Williams, who currently serves as the CEO.

AO has a large community supporting the project, with sectors from DeFi to GameFi to AI all building on AO as well as a 35 million dollar venture fund specifically funding projects built on AO.

Further details on the community surrounding AO can be read in the tweet below:

Defi_Mochi on X: “Looking to get into $AO @aoTheComputer ? Here’s the @aoTheComputer complete ecosystem project list Community / Support 1. @fwdresearch — Research and incubator 2. @Weavers_Org Dev and community 3. @aoTheVentures Ecosystem venture 4. @betteridea_dev Developer ecosystem 5.” /

Financial Applications on AO:

The possibilities of integrating AI models into smart contracts on AO leave potential to build some of the most advanced autonomous DeFi applications (AgentFi). The biggest use cases will be improving core infrastructure, implementing AgentFi for DeFi users and the possibilities of TradFi adoption.

Core Infrastructure:

Firstly, AI can be leveraged into core infrastructure to enhance key applications such as stablecoins, DEX’s and money markets. While these DApps are currently built on other chains and most likely cannot be moved to AO due to funding, community and migration issues, the next generation of financial DApps are incentivised to build on AO. Some key examples of how AO can improve core infrastructure are:

1. Machine learning models that analyse transaction data before processing and are trained to identify and correct anomalies in transaction data to increase the overall accuracy of transactions.
2. Forecasting periods of low liquidity and adjusting market mechanisms through fee modification or market-making strategies. This method can also be used to optimise operations for cross-chain bridges, stablecoin collateralization and providing faster matching algorithms for DEX’s.
3. AI models can stress test extreme market conditions to find potential liquidity issues before they occur to build a more resilient market.

Agent Fi:

Autonomous agents are programmed entities that operate on predefined algorithms and use machine learning and predictive analytics to make decisions when analysing real-time market data. Currently, the data available across DeFi that can be utilised is quantitative only. Furthermore, DeFi bots require significant oversight to manage as they execute strategies based on preconfigured parameters that can become useless due to changing market conditions. These bots lack the sophistication to dissect market news or any other qualitative data which is where autonomous agents shine. Autonomous agents have the potential to dissect reports, speech and news and predict and capitalise on the resulting price changes in accordance with their strategies. These strategies which will be determined by the LLM’s on AO will have the potential to create limitless 24/7 trading agents that can learn and adapt to fluid market conditions, simulating the strategies of the best investors with minimal manual labour.

A few examples of financial use cases for AgentFi include DCA Agents, Autonomous Portfolio Managers and Risk Balancing Agents. Decentralised hedge funds can also now be created with immutable strategies, with users being able to safely participate without having to trust the person or centralised entity managing it, due to the verifiability of AO. As mentioned earlier, developers can also select the most compatible EVM’s and schedulers to build the best version of their product, further incentivising creation on AO rather than another computation platform.

TradFi Adoption:

Traditional Finance (TradFi) firms could also explore the integration of decentralised financial agents on AO due to several benefits such as:

1. AO’s immutable and tamper-resistant ledger ensures that all transactions and operations are permanently recorded and cannot be altered, providing a higher level of security compared to centralised databases.
2. In a decentralised environment like AO, the reliance on intermediaries and counterparty risk that is typical in TradFi is lowered significantly. In TradFi, transactions are dependent on the trustworthiness and solvency of intermediaries like market makers and AO is a solution that can offer a more transparent and competitive investment environment.
3. The most likely form of adoption could come from startup funds that are eager to acquire capital to grow and compete with industry giants. As mentioned earlier, the verifiability of AO allows users to invest without ever meeting managers and thus makes it easier for emerging funds to compete. This accessibility provides a substantial advantage and with the current trend of TradFi tokenizing real-world assets, there is evidence to suggest TradFi’s openness to embrace DeFi solutions when there is a clear financial incentive, which AO does provide.

Training AI models on AO:

Leveraging AO’s unlimited computation, AO can support the training and deployment of high-performance AI large language models (LLM’s) to operate directly within smart contracts. AO leverages Arweave’s ability to provide verifiability through permanent data storage which is crucial for AI training as the integrity of training data must be maintained over time.

Centralised AI models raise concerns regarding data privacy, accessibility of cloud computing and centralised decision-making processes. The main challenges with centralised AI models are as follows:

1. Centralised AI systems lack transparency over data usage as large companies utilise data with zero obligation of transparency to users.
2. As AI companies require computing power they must cooperate with cloud service giants creating a monopoly on cloud services.

Decentralised AI development ensures AI is both secure and private. This increases competition and reduces centralised control from big companies.

AO is the optimal computation platform to train AI models as it is the only option that is decentralised, verifiable, scalable and economically competitive.

There are no other blockchains that are both scalable and verifiable because AO can only offer verifiability due to its base layer Arweave. This technology is replicable, but without a significant value add it will be difficult to overcome AO’s first mover advantage, especially as Arweave was and still is the distinct permanent storage solution.

The main hurdle that AO faces is the privacy concerns that will result from computations being stored on Arweave from users, particularly regarding sensitive or confidential information. This relates to industries that already have some scepticism regarding the integration of AI such as the medical and financial sectors. The solution to this issue is Fully Homomorphic Encryption (FHE), a solution which is already in the works, as seen below:

​​https://twitter.com/samecwilliams/status/1802166861976535310

FHE is a specific data encryption method that enables users to perform computations on encrypted data without revealing any of the original raw data. FHE allows complex calculations to be carried out on encrypted data and generates an encrypted result that when decrypted, matches the result of the original operations. This ensures data can remain secure and private while still being accessible. This is a necessity for industries with privacy concerns but have significant productivity advancements that can be achieved with AI.

A key weakness at this time is a lack of graphical user interfaces on AOS which will present a challenge for Web2 developers that are more accustomed to more visually intuitive development environments. AOS is the operating system that manages how decentralised applications operate on AO. AOS provides the tools and interfaces for developers to deploy and interact with applications running on AO. Similar to how Windows manages applications on a computer, AOS manages the decentralised applications running on the AO network. While AOS simplifies the process of developing on AO through the programming language LUA, which only takes 15 minutes for a developer to learn, it still is not an optimal environment and can hinder the transition to building on AO with solutions like ChatGPT enterprise satisfying privacy concerns in the short-term.

Ultimately, given the AI market is predicted to grow to 2.575 trillion by 2032, AO has the potential to capture a noticeable share of the computing market as it remains the superior option with further updates on the way. Given every computation will be directly stored on Arweave, the growth of use of the AO network will have a proportional impact on the demand for AR. This is why gaining exposure to AR can be seen as investing in the growth of DeFi, AI and NFT’s all at once.

Other Use Cases:

Some other notable use cases that can be built on AO using autonomous agents are tokenized games, AI chatbots, decentralised recommendation systems and indexes for social media. A notable example is Odyssey, a web3 social video sharing platform with 7 million monthly active users and a fully decentralised peramweb application.

Key Risks:

1. Testnet and Mainnet Challenges

AO’s current testnet phase before the mainnet launch may reveal technical difficulties as it is further stress tested. Furthermore, there are always unknowns which can be exploited with a new launch and AO needs time to be battle-hardened before users can be fully confident in AO’s security.

2. AO Token Integration

The primary role of the AR token is to ensure data storage on the Arweave network. This includes paying for data storage and rewarding miners for adding blocks to the network. The AO token in contrast will be designed to handle computational and communication functions within AO. This means managing the messaging between different processes on the network. AO token incentives will likely provide incentive mechanisms for nodes (computers) that participate in the decentralised computation process. However, the AR token will still gain exposure to AO’s value, even with the creation of an AO token as all computations will still be stored on Arweave’s network permanently. This means all usage of AO’s network will directly correlate to an increased demand on Arweave’s network and demand for the AR token. The risk in this scenario however is that the token economic design of the AO token is not optimal in incentivizing participation of nodes and due to the lack of consensus currently about the design, it is a clear uncertainty that could torpedo the project.

Conclusion:

Ultimately, AR derives value from both Arweave (the base storage layer) and AO (the computing layer). Arweave’s dominance in permanent decentralised data storage ensures it can capitalise on ensuring the long-term security of NFT’s, a market predicted to triple by 2030. Furthermore, AO acts as an optimal computing layer for the development of AI models and Autonomous Financial products. This positions AO to capture the rapid expansion of the AI market, placing AR at the intersection of multiple different growth trajectories. While still in early stages, Arweave and AO can face both financial and technical issues, however their strong record of success and strategic positioning signals a promising outlook in use cases and network growth over the next 5 years.

Works Cited

DAO, Perma. “Arweave Market Intel.” Medium, 14 May 2024, medium.com/@perma_dao/arweave-market-intel-6c5e93e639a2. Accessed 9 June 2024.

— -. “Arweave: The Ultimate Solution for NFT On-Chain Storage.” Medium, 7 Nov. 2023, medium.com/@perma_dao/arweave-the-ultimate-solution-for-nft-on-chain-storage-6152d603ec03.

— -. “Can Arweave AO’s AgentFi Dispel the Illusion of Crypto X AI?” Medium, 14 Apr. 2024, medium.com/@perma_dao/can-arweave-aos-agentfi-dispel-the-illusion-of-crypto-x-ai-a61003abccc5. Accessed 18 June 2024.

— -. “Is AO’s Financial Ecosystem Ready to Take Off?” Medium, 3 May 2024, medium.com/@perma_dao/is-aos-financial-ecosystem-ready-to-take-off-0acde4b02947. Accessed 13 June 2024.

— -. “The Future of AO — an Application Consensus Interaction System with Infinite Scalibility, a Event….” Medium, 29 Mar. 2024, medium.com/@perma_dao/the-future-of-ao-an-application-consensus-interaction-system-with-infinite-scalibility-a-event-3f57dd8aa760. Accessed 9 June 2024.

— -. “Wealth Creation Effect Emerges!” Medium, 16 May 2024, medium.com/@perma_dao/wealth-creation-effect-emerges-ec8245aac886. Accessed 18 June 2024.

Hourani, Mustafa. “Explaining the Recent Rise of Fully Homomorphic Encryption in the Blockchain Industry.” Medium, 29 Oct. 2023, medium.com/@mustafa.hourani/explaining-the-recent-rise-of-fully-homomorphic-encryption-in-the-blockchain-industry-c7081fa05458.

Mochi, Defi. “X.com.” X (Formerly Twitter), 8 June 2024, x.com/defi_mochi/status/1799304617186275663. Accessed 9 June 2024.

Research, Sistine. “Arweave Research Report.” Arweave Research Report, sistine.ai/research/arweave-research-report. Accessed 18 June 2024.

Storage, NFT . “NFT.Storage.” Nft.storage, nft.storage/blog/nft-storage-endowment-model. Accessed 19 June 2024.

Williams, Sam, et al. Arweave: A Protocol for Economically Sustainable Information Permanence.

YBB. “From Storing the Past to Computing the Future: The Hyper Parallel Computer of AO.” YBB Capital, 3 Apr. 2024, medium.com/ybbcapital/from-storing-the-past-to-computing-the-future-the-hyper-parallel-computer-of-ao-3335a3f60ab3#:~:text=Simply%20put%2C%20AO%20can%20be. Accessed 9 June 2024.